Premium
Configuration and hindcast quality assessment of a Brazilian global sub‐seasonal prediction system
Author(s) -
Guimarães Bruno S.,
Coelho Caio A. S.,
Woolnough Steven J.,
Kubota Paulo Y.,
Bastarz Carlos F.,
Figueroa Silvio N.,
Bonatti José P.,
Souza Dayana C.
Publication year - 2020
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1002/qj.3725
Subject(s) - hindcast , climatology , mean squared error , environmental science , empirical orthogonal functions , anomaly (physics) , forecast skill , madden–julian oscillation , initialization , precipitation , meteorology , mathematics , statistics , convection , geology , computer science , geography , physics , programming language , condensed matter physics
This article presents the Centre for Weather Forecast and Climate Studies (CPTEC) developments for configuring a global sub‐seasonal prediction system and assessing its ability in producing retrospective predictions (hindcasts) for meteorological conditions of the following 4 weeks. Six Brazilian Global Atmospheric Model version 1.2 (BAM‐1.2) configurations were tested in terms of vertical resolution, deep convection and boundary‐layer parametrizations, as well as soil moisture initialization. The aim was to identify the configuration with best performance when predicting weekly accumulated precipitation, weekly mean 2 m temperature (T2M) and the Madden–Julian Oscillation (MJO) daily evolution. Hindcasts assessment was performed for 12 extended austral summers (November–March, 1999/2000– 2010/2011) with two start dates for each month for the six configurations and two ensemble approaches. The first approach, referred to as Multiple Configurations Ensemble (MCEN), was formed of one ensemble member from each of the six configurations. The second, referred to as Initial Condition Ensemble (ICEN), was composed of six ensemble members produced with the chosen configuration as the best using an empirical orthogonal function (EOF) perturbation methodology. The chosen configuration presented high correlation and low root‐mean‐squared error (RMSE) for precipitation and T2M anomaly predictions at the first week and these indices degraded as lead time increased, maintaining moderate performance up to week‐4 over the tropical Pacific and northern South America. For MJO predictions, this configuration crossed the 0.5 bivariate correlation threshold in 18 days. The ensemble approaches improved the correlation and RMSE of precipitation and T2M anomalies. ICEN improved precipitation and T2M predictions performance over eastern South America at week‐3 and over northern South America at week‐4. Improvements were also noticed for MJO predictions. The time to cross the above‐mentioned threshold increased to 21 days for MCEN and to 20 days for ICEN.